IMAGE FILTERING AND PERFORMANCE ANALYSIS UNDER VARIOUS TYPE OF NOISE

pp 9‐ 12

1Neesha Sharma, 2Gaurav Sharma

1M.Tech scholar, Department of Computer Science Engineering,  SGVU, Jaipur

neeshashish2907@gmail.com

2Assistant Professor, Department of Computer Science Engineering, SGVU, Jaipur

gauravkr.sharma@mygyanvihar.com

 

Abstract: In this paper we are we have recovered the image from Gaussian and pepper & salt noise. Here we have a colored image. We are converting the colored image into grayscale image. In the gray scaled image we are adding Gaussian noise and remove by winner filter and median filter and similarly in the same image we add paper and salt noise and remove the noise by the means of same filters and finally check their performances. The performance criteria is mean square error.

Keywords:  Image Processing, Gaussian Noise, Pepper & Salt Noise, Adaptive Filtering, Median Filters.

 

INTRODUCTION

Image processing is an important task in several applications like satellite imaging, medical diagnosis, picture editing and agriculture etc. In the early age of image processing we have analog image processing methods are there but in the present age of image processing we are using digital image processing. Some very good tools of transformation like wavelet [1] transform is also there even machine learning and deep learning also used in the image processing.

Various type of noise always affect image and result is image blurred or information loss in the image. There are several filters to recover the image [3-6] from various type of noise. Some filters are static in behavior and some are adaptive or intelligent in nature.

In this paper we are we have recovered the image from Gaussian and pepper & salt noise. Here we have a colored image. We are converting the colored image into grayscale image. In the gray scaled image we are adding Gaussian noise and remove by winner filter and median filter and similarly in the same image we add paper and salt noise and remove same filters and finally check their performances.

  1. NOISE IN IMAGE

When image is transmitted over the channel noise is added to the image or sometimes it is blurred because of camera quality. So it is important to restore the image by removing the various noises from the image. There are several types of noises like Gaussian noise, Poisson noise, pepper & salt noise etc. To remove these noises we have several types of linear & non linear filters like winner filter [2], kalman filter, adaptive median filter[6,7] etc.

  • WINNER FILTER

Winner filters are characterized as

  1. It is assumed that the type of noise is AWGN. Noise is characterized as a stationary and linear process.
  2. Winner filter is realized as causal filter.
  3. Performance criteria of the winner filter is minimum mean square error (MMSE).

 

 

 

 

Estimation of noisy image

Where ‘g’  is the deconvolution filter.

So winner filter is the inverse filter.

III. RECOVERY OF IMAGE UNDER VARIOUS NOISE CONDITIONS AND VARIOUS FILTER TYPES.

In this analysis we have a color image, first we have convert this image into gray scale image. The original image is showing in Fig.1 and grayscale image is showing in Fig.2

In the first case we added the pepper & salt noise in the image which is shown in Fig.3

When this noise is filtered by winner filter is showing in Fig 4 and output of median filter is showing in Fig 5.

In Case 2 we add the gaussian noise in the gray scale image shown in Fig.2 and filtered by again the same filters. The Gaussian noise added image is showing in Fig. 6.

Fig.7& Fig.8 are showing the output of winner filter and median filter respectively.

By figure 4 and 5 we can that median filters better removes the pepper & salt noise from the image. Similarly in case of Gaussian noise it is difficult to remove the noise completely by both of the filters. But winner filter having good performance over median filter. The MMSE of all possible cases is present in table.1

Table.1 summarizing the whole picture of performance of both filters under both kind of noise.

CONCLUSION

In the image processing , denoising of the image is an important task. By this experimental work we can say that for various type of noise various filters are required. One type of filter can perform in various noisy environment. median filters better removes the pepper & salt noise from the image. Similarly in case of Gaussian noise it is difficult to remove the noise completely by both of the filters. But winner filter having good performance over median filter.

REFERENCES

  1. Sang-Hee Yoo, Jae-Wook Jeon and Jung-Hyun Hwang, “Spatial-temporal noise reduction filter for image devices,” 2008 International Conference on Control, Automation and Systems, Seoul, 2008, pp. 982-987.
  2. Arazm, A. Sahab and M. F. Kazemi, “Noise reduction of SEM images using adaptive Wiener filter,” 2017 IEEE International Conference on Cybernetics and Computational Intelligence (CyberneticsCom), Phuket, 2017, pp. 50-55.
  3. Choi and J. Jeong, “Despeckling Images Using a Preprocessing Filter and Discrete Wavelet Transform-Based Noise Reduction Techniques,” in IEEE Sensors Journal, vol. 18, no. 8, pp. 3131-3139, April15, 15 2018.
  4. Tasnim, M. M. H. Shuvo and S. Hasan, “Study of speckle noise reduction from ultrasound B-mode images using different filtering techniques,” 2017 4th International Conference on Advances in Electrical Engineering (ICAEE), Dhaka, 2017, pp. 229-234.
  5. Vijayalakshmi, S. Aparna, G. Gopal and W. J. Hans, “Handwritten character recognition using diagonal-based feature extraction,” 2017 International Conference on Wireless Communications, Signal Processing and Networking (WiSPNET), Chennai, 2017, pp. 1178-1181.
  6. Zhang, X. Li and C. Zhang, “Modified Adaptive Median Filtering,” 2018 International Conference on IntelligentTransportation, Big Data & Smart City (ICITBS), Xiamen, 2018, pp. 262-265
  7. Zhang et al., “SAR image change detection method based on shearlet transform,” 2017 Progress in Electromagnetics Research Symposium – Fall (PIERS – FALL), Singapore, 2017, pp. 1223-1229.
  8. Jianhua Pang, S. Zhang and Wencang Bai, “A novel framework for enhancement of the low lighting video,” 2017 IEEE Symposium on Computers and Communications (ISCC), Heraklion, 2017, pp. 1366-1371.